Seller activity in a virtual marketplace
First Monday

Seller activity in a virtual marketplace by David A. Huffaker, Matthew Simmons, Eytan Bakshy, and Lada A. Adamic



Abstract
As virtual goods continue to proliferate in online worlds, understanding their production, consumption and distribution remains exciting for scholars, technology companies and policy makers alike. We present a descriptive study of the activities of successful sellers in Second Life, a 3D virtual world that allows users to create their content and even to make money by selling it to other users. We combine user log analysis, network analysis and content analysis to examine cycles in trading volume, market segmentation and specialization, geographic concentration and the impact of social capital on economic success, revealing important insights regarding virtual markets, as well as differences between the very top sellers and those making a more modest income.

Contents

Introduction
Background
Methods
Results
Discussion and conclusions

 


 

Introduction

The economies of virtual worlds have been of interest to scholars because they provide opportunities to understand macroeconomic behavior, market capital and business potential (Castronova, 2001; Malaby, 2006). Virtual goods, which can be defined as “characters, items, currencies and tokens” found in virtual worlds and games, are estimated to be a US$2 billion dollar industry (Lehdonvirta, 2009). Popular media has highlighted the growing popularity of virtual goods in a variety online communities and the accompanying rush by venture capitalists to explore another avenue for online revenue (Cain Miller and Stone, 2009; Levy and Satariano, 2010). Additionally, scholars are forecasting the continued growth in consumption of virtual goods by younger Internet users who enjoy the entertainment, competitiveness and social payoffs associated with virtual goods (Welch, 2010). Understanding what these virtual marketplaces look like — and more importantly, what successful sellers look like — remains interesting to scholars, technology companies and policy makers alike. For researchers, they provide insights on economic theory and market behavior; for technology companies, they offer new strategies for monetizing online spaces; for policy makers, they provide evidence for regulation or deregulation.

Few researchers have had access to actual data on virtual economies that can support empirical claims about how virtual goods proliferate in the marketplace or characteristics of the most successful sellers and market segments. The purpose of this study is to describe the economic environment of a virtual marketplace and identify attributes of the most successful sellers. We rely on a unique set of data from Second Life, a 3D virtual world that allows users to create their own avatars and objects, socialize and interact with other users, and explore a vast landscape. Second Life is a particularly interesting context of study because the majority of the content is user–generated (Ondrejka, 2004), and more than half of its residents have created commodities varying from vehicles and buildings to clothes and games (Ondrejka, 2008). Although new users receive a small amount of money when they enter the world, they are able to increase their wealth by using real money to buy Linden Dollars. The sheer volume of revenue that Second Life generates is impressive; according to Second Life’s founders, hundreds of thousands of residents will engage in 10 million transactions, amounting to over US$15 million dollars in a single month (Ondrejka, 2008).

Given the potential for a virtual economy such as the one thriving in Second Life to provide important insights into market behavior, it is important to ask: What does economic activity in a virtual world look like? What does seller activity focus on, and what are the primary characteristics of the most successful sellers? In order to answer these questions, we utilize a multi–method approach that combines the analysis of user logs, the transaction, chat and social networks of the top sellers, and a qualitative content analysis of the types of online groups they belong to. This allows us to identify patterns of production, consumption, distribution and exchange (Marx, 2003), the interplay of material, cultural and social capital (Bourdieu, 2001), and the primary sectors or market segments that constitute virtual economies.

 

++++++++++

Background

The majority of research on virtual economies focuses on the shared features with ‘real–world’ economies. Castronova’s (2001) study of an early massive multiplayer online role–playing game (MMORPG) paralleled wealth accumulation among Norrath players with the gross national product of developing countries, and highlights how the supply and demand of items needed to enjoy the world create “bazaar”–like exchanges between players [1]. Later, he predicted a huge growth in virtual world popularity and virtual market revenue, and potential macroeconomic impact for real–world economies (Castronova, 2002). Most recently, Castronova and colleagues showed that aggregate–level economic behavior in Everquest II, a popular MMORPG, follows similar patterns to those seen in real–world economies, and that the majority of virtual goods exchanged represent real–world objects, including raw materials, food, clothing, furnishings, as well as items unique to the game such as weapons, scrolls and recipes for new spells or trade skills (Castronova, et al., 2009). Together, this research reveals important similarities between real and virtual economies in terms of revenue patterns, exchange processes and types of products being exchanged, even when these phenomena are partially constrained by game developers.

Scholars of marketing and management have also been interested in virtual economies, often focusing on how to best leverage traditional economic and marketing principles in this new computer–mediated space. For example, some research argues that social interactions drive economic activity, and that marketers or managers must embed virtual market strategies within the social processes that emerge in a virtual community (Balasubramanian and Mahajan, 2001). Rothaermel and Sugiyama (2001) suggested that large virtual communities have the potential to attract more vendors and advertisers, resulting in more transactions and greater revenue, but this is contingent on the ability of a community to scale in a way that keeps users interconnected and socially active.

Other work has focused on the motivations of buyers and attributes of virtual items that drive purchasing behavior. In a review of over a dozen virtual worlds including EVE Online, Second Life and World of Warcraft, and interviews with several virtual world developers, Lehdonvirta (2009) argues that functional attributes (e.g., items that provide performance advantages, new behaviors, or more convenience), hedonic attributes (e.g., aesthetics, fashion), or social attributes (i.e., being stylish or showing off to other players, rarity of items) all drive users to purchase virtual goods. In a focus group study of 24 players of Second Life, Everquest II and World of Warcraft, Guo and Barnes found that social influence, resource scarcity, real–world monetary budget, and performance expectancy all contribute to buying virtual goods. Although this preliminary work provides a foundation for understanding virtual markets, little research has focused on the seller perspective.

It is important to note that Second Life is distinct from MMOGs such as World of Warcraft, Everquest II, or even Facebook apps in that the majority of the virtual goods are user–generated. While Linden Lab has provided customizable avatars, land, a flexible scripting language and a few laws of virtual physics, the cities, resorts, dance clubs, fashion scene and games have all been created by the actual users. Almost immediately, residents began creating clothing and accoutrements, and soon entrepreneurs and speculators began building infrastructure to support the economy including boutique stores, shopping malls and real estate for expansion. Sellers also began to distinguish themselves by diversifying products, using advertising outlets, and providing immersive shopping experiences (Ondrejka, 2008).

Second Life’s content reflects many of the industries of the real–world economy, including manufacturing and retail outlets, infrastructure such as roads and buildings (including public services such as hospitals and schools), and a variety of services including marketing, attorneys, insurance companies and other financial services. However, the economy is also unique in that the production, reproduction and distribution costs traditionally associated with distribution of goods and services are dramatically reduced (Malaby, 2006). Residents can produce their goods from scratch, or modify existing goods, but they are not required to use base materials (i.e., buy concrete to build roads, or steel to create buildings).

Second, residents can duplicate their products at zero–cost, enjoying little marginal production costs and immediate economies of scale. Third, they do not require infrastructure such as electricity or water/sewage in order to operate a store, although they are limited by real estate. In other words, although Second Life has the capability to replicate the industries and markets of the physical world, the creators do not have all the constraints that real–world producers face. So a central goal of this research is to understand how supply and distribution is affected when the costs of production are no longer a major boundary.

With little beyond real estate needed in the way of physical infrastructure, sellers can focus their efforts on creating innovative goods and attracting customers to them. Groups, clubs and events are designed to support business activity and become important money–making ventures on their own. They draw in crowds and represent the most traffic within Second Life, and are used to sell goods or services and feature new products (Ondrejka, 2008). A resident can concurrently be a member of up to 25 different groups in Second Life, which can be organized around common interests (e.g., poetry readings or vampires), or places (e.g., dance clubs or islands), but often serve as a way to broadcast advertisements for new products or events (Boellstorff, 2008). Therefore a specific goal of this research is to examine the relationship of groups and successful sellers.

There are several economic and marketing principles that are relevant in the context of virtual marketplaces. Beginning in the late nineteenth century, economists focused on the dynamic nature of prosperity and depression in economic systems in an attempt to forecast their depth and duration. These expansions and contractions are referred to as business cycles, and all economies are subject to these peaks and valleys, which can be short–term (e.g., monthly downturn, seasonally) or long–term (e.g., yearly or every several years) (Knoop, 2004). Business cycles even occur at a micro–level. For example, retail markets typically enjoy higher demand on weekends, and often promote markdowns to increase sales even further (Warner and Barsky, 1995). While some scholars argue that the impact of weekend effect would be reduced in online retail markets (Scholten, et al., 2009), we still expect to see some fluctuation in the amount of revenue and transaction volume witnessed in this virtual economy.

Second, because consumer demand is fickle and complex (Peter and Olson, 1999), markets compensate through market segmentation and specialization, in which products are customized to meet the geographic, demographic, psychographic (e.g., lifestyle, personality or social class), or behavioral (e.g., attitudes, readiness, loyalty) attributes of potential customers (Kotler and Armstrong, 2008). Sellers must diversify their product line to appeal to market demand, compete with other sellers and carve out their own niche. Scholars argue that market segmentation and specialization is especially useful online, where a much richer set of data regarding consumer behavior is available, and items can be personalized to a greater degree (Wedel and Kamakura, 2000). Therefore, we expect sellers in Second Life to demonstrate some diversification in their products, including the ways in which they are distributed.

Third, scholars have noted that geographic concentration and clustering of marketplaces are a prominent feature of almost all national, regional and metropolitan economies (Porter, 2000). Shopping malls and strip malls benefit from providing consumers with the convenience of location (Reimers and Clulow, 2004). Studies show that competing retail stores selling similar items benefit from being co–located, and attract more consumers who are uncertain about what they want or seek a low price (Konishi, 2005). Although we recognize the removal of spatial constraints in online settings, we still expect to see some geographic concentration in the virtual marketplace because it provides the similar conveniences for potential customers who want to see all the products available in a particular genre, or who are simply browsing.

Fourth, economic activity is often embedded in social practices. Bourdieu (2001) highlights the interplay between different forms of capital: economic (e.g., resources and assets), social (e.g., group membership and social support) and cultural (e.g., knowledge, skills, status). He suggests that each these forms can be exchanged and transformed to another form of capital. Malaby (2006) extends this theoretical framework to virtual worlds, arguing that users can exploit commodities and currency (i.e., economic capital) through their connections with others (i.e., social capital) and their credentials (i.e., cultural capital). Therefore, we expect to see examples of social capital and social relationships associated with top sellers.

 

++++++++++

Methods

Sample. Linden Lab, the creators of Second Life, provided a collection of anonymized user log data containing: (a) transaction logs, which identify when an item is transferred from one user to the next, the location and date of the exchange and the amount that was transferred (including zero–cost items); (b) chat logs, which identify when one user sends a chat message to another through the in–game chat system, and the amount of characters in the message on a daily basis; (c) social logs, which identify when users are friends (i.e., a ‘buddy list’) on a weekly basis; and (d) the list of groups that each was is a member of on a weekly basis.

We focused our analysis on a subset of data representing April 2009, for which transaction, chat, social and group data were available concurrently. The amount of user activity in this single month is vast: over 66,237,390 million transactions; 17,753,736 million sent chats, 43,482,786 million social buddies and 829,458 groups represented.

Rather than focus on the entire set of users, we extracted two subsets of users that represent the successful sellers in Second Life. First, we calculated the total revenue for each user in April, and then selected the top 100 users who generated the most revenue (i.e., “Top 100 Sellers”). Second, we randomly selected a sample of users around the mean revenue value of the entire population. The majority of users make very little revenue, and the mean value actually represents the 93–94 percent quantile of sellers. Therefore, in order to get a more representative sample of ‘typical’ sellers we randomly sampled 100 sellers from a selection of 16,351 users within +/– 3 percent of the mean (i.e., “Boutique 100 Sellers”).

Transaction logs. Transactions refer to events in which one user transfers an item to another user in exchange for Linden Dollars. For each transaction that occurred in April 2009, the user ids for the seller and buyer of any item are recorded. Unless indicated, we only extract transactions that represent payments between avatars, object sales (i.e., from a store vendor), inventory exchange between avatars, and when an object (e.g., a machine) pays out. In the case of objects, the owners or creators of these objects are identified as the seller. The seller and buyer link is also used to create network graphs.

For each transaction, the amount of Linden Dollars that is exchanged is collected. The amount of Linden Dollars can also be zero, but we filter out zero–cost items for all analyses unless identified. The actual two–dimensional x,y coordinates where each transaction took place in the virtual world is collected. The time and date (in seconds) when each transaction took place is collected. These UNIX timestamps were converted to hourly and daily aggregates for data analysis.

Chat logs. Second Life offers an in–game chat system in which users can send private messages to one another. Our data set does not include any chat content, but rather just the number of messages and total number of characters exchanged on each day. This data is used to create directed and weighted network chat graphs.

Social network logs. Second Life provides users with the ability to select specific users as ‘friends’. The social log lists all friendship links current through April 2009. The friend–friend link is also used to create network graphs.

Group types. Second Life users are able to be members of up to 25 groups at one time. These groups range from lifestyle or hobby groups, to fan clubs or other entertainment uses, or even land, finance and educational topics. The user logs contain both group names and charters, which describe the purpose of the group. Using a grounded theory approach (Strauss and Corbin, 1990), we reviewed a random selection of group names and charters and constructed a collection of group ‘types’ or genres to represent several basic sectors. Next, we randomly selected 20 percent of the groups that each seller (both Top 100 and Boutique 100) belonged to in April 2009. An interrater reliability analysis using a second coder and Cohen’s Kappa was performed. The interrater reliability showed very good agreement (κ = .94, ρ <.001).

 

++++++++++

Results

The Second Life (SL) economy demonstrates an extreme degree of market centralization, as measured by the proportion of revenue taken in by each account. 272,524 users engaged in at least one transaction during the month, but a small percentage of these users accounts for a large chunk of total revenue. The Gini Index (Sen, 1976; Shalit, 1985), which can range from 0 (all sellers’ revenues are exactly equal) to 1.0 (all the sales are made by a single seller), holds remarkably steady at 0.963±0.008 over a roughly two–year period (from July 2007 to May 2009). The Top 100 Sellers make up 40.98 percent of the total revenue in any given month (by contrast, the Boutique 100 users only represent .04 percent of the total revenue). In comparison, the top 100 largest manufacturing corporations in 1997 comprised 32 percent of the “value added” by manufacturers in the United States (White, 2002).

It is therefore of interest to trace and characterize the activities of these top sellers, as they reflect a large portion of economic activity in SL. However, the Top 100 Sellers are by no means a constant. If we examine their population in April of 2009, 71 of them were in the Top 100 a month earlier, only 32 of them were in the Top 100 a year earlier, in April of 2008, and only 14 of them were in the Top 100 in August of 2007. Therefore, although top sellers tend to dominate, there are always opportunities for new top sellers to emerge. Furthermore, we find it instructive to analyze the activity of top sellers for a given month because it is their aggregate activity — rather than the activity of any one (replaceable) individual — that is of interest.

Table 1 shows the means and standard deviations for transaction volume, as well as the number of friends that the users had during the month, and the number of recipients of chat message. Again, the Top 100 Sellers demonstrate much higher transaction volume than either the user population as a whole or even the boutique seller comparison group. They also have more friends and chat partners. The Boutique 100 Sellers were selected to have revenue approximately equal to the population average, but due to the high skew in revenue distribution that actually places them in the 93rd revenue quantile. They also have more friends and chat buddies than the average Second Life user. This marks two important points immediately: we find that (1) a small subset of users are dominating virtual sales both in terms of revenue and the volume of exchange; and (2) top sellers show a richer contact network than the average user.

 

Table 1: Summary of the means and standard deviations of transaction, friendship and chat attributes for top sellers and all users.
Notes: a. N = 272,524; b. N = 4,282,715; c. N = 450,956.
 Means
(standard deviation)
VariableAll usersTop 100 SellersBoutique Sellers
Number of transactions169.91
(2,574.81)a
39,719.62
(56,839.80)
297.53
(493.36)
Number of friends10.15
(37.61)b
102.09
(163.31)
98.31
(96.28)
Number of chat buddies36.37
(280.31)c
1,366.54
(3,606.88)
181.03
(433.73)

 

In addition to the raw numbers involved in economic activity among top sellers, we are interested in the patterns or cycles involved in a given month, where these transactions take place, and the impact of social interaction and group membership. We discuss these results in the next sections.

Daily cycles. The idea that economies experience expansions and contractions at many depths and durations is well known, and we expected to see fluctuations in the economic activity of Second Life as well. In order to examine this, we extract the revenue and transaction volume with hourly resolution for each day in April 2009. Because some sellers generate very high revenue or transaction volume — which inflates the average — we rely on a more representative median value. Figure 1 shows the median total hourly revenue and Figure 2 shows the median total number of transactions. At the hourly–level, we see periodic fluctuations in both revenue generation and the number of transactions, and similar fluctuations in the for both Top 100 and Boutique sellers. However, we find no cyclical or non–cyclical pattern in the revenue or transaction volume at the daily level. In particular, there are no differences between weekday and weekend sales as occur for brick–and–mortar stores.

 

Figure 1: Median hourly total revenue for Top 100 and Boutique Sellers normalized by median for entire month of April 2009. Hour 1 begins at midnight on April 1
Figure 1: Median hourly total revenue for Top 100 and Boutique Sellers normalized by median for entire month of April 2009. Hour 1 begins at midnight on April 1.

 

 

Figure 2: Median hourly transaction volume for Top 100 and Boutique Sellers over a month
Figure 2: Median hourly transaction volume for Top 100 and Boutique Sellers over a month.

 

Geographic concentration. Because a concentration of businesses is positively related to economic activity in the physical world, we expected to see some concentration in the economic activity of our seller groups. In order to test this hypothesis, we extracted the x,y coordinates of every transaction that occurs in April. We plot the coordinates for each transaction to highlight how often the transactions occur. This allows us to examine whether transactions occur in common spaces such as a shopping center or commercial hub, or if selling is distributed all over the Second Life world.

Although the coordinates range from 0–1400 for both the x– and y–axes, we find that almost all transactions occur in the upper right quadrant (x–axis: 400–1200; y–axis: 800–1400) of the Second Life map. This area is mainly made up of several main lands and a large collection of islands covering a variety of content, from dance clubs and gaming clubs, to rural landscapes and housing resorts. We find that these areas are densely populated, making it a prime substrate for economic activity.

As shown in Figure 3, the Top 100 Sellers conduct transactions over a large portion of the region. However, the 3,924,133 transactions that took place in April occur in only 5,654 locations, and there are a couple locations (highlighted in red) in which a high volume of transactions takes place (i.e., over 400,000). A closer inspection of the top ten areas with the most revenue is divided between the main lands and the individual islands, and are a mixture of gaming zones and retail stores. Some of these zones encompass a larger land that offers visitors places to relax and socialize (e.g., gardens and clubs), along with retail options (e.g., stores selling clothing and gestures).

 

Figure 3: Transaction locations for Top 100 and Boutique Sellers overlaid on a map of Second Life. Blue represents the virtual world's ocean and the grey masses are the land
Figure 3: Transaction locations for Top 100 and Boutique Sellers overlaid on a map of Second Life. Blue represents the virtual world’s ocean and the grey masses are the land.

 

We see a similar pattern with respect to Boutique Sellers. The top ten highest revenue–generating locations were almost all individual islands, and included several rental zones along with retails shops. In these cases, we find that individual sellers use an island to sell unique items (e.g., a teddy bear workshop) or rent land and housing that they have built.

We also compare how concentrated seller activity is across the different groups by measuring what proportion of transactions occurs within a radius of 50 of one another. If transactions were distributed entirely uniformly, the proportion would be the area of a circle of radius 50 divided by the area of the populated region, conservatively (1100–400)*(1300–900). We would then expect only two percent of transactions to occur within a distance of 50 of one another. Because Second Life is not uniformly developed across geography, the actual percentage across all transaction is 7.96 percent. The percentage is a bit higher, at 12.13 percent for Boutique sellers, indicating that although Boutique sellers individually may be operating in a small area, our random sample of boutique sellers is geographically dispersed. However, for the Top 100 Sellers, a full 19.45 percent of transactions occur within such a short distance. We conclude that although top sellers cover a large geographical area due to the sheer volume of their transactions, the majority of their activity is collectively geographically more highly concentrated than that of the Boutique sellers. This is consistent with their revenues stemming from a single market segment, as will be discussed in the next section.

The social capital of economic activity. Finally, because economists and social theorists have noted a relationship between economic capital and social capital, we expected that our seller groups would interact socially in addition to conducting business transactions. Therefore we superimpose two other networks on the transaction network within our seller groups: communication in the form of chat and online friendship in the form of the SL friend graph. Users can add one another as friends, which enables them to see when they are online (by default) to coordinate chat, and also optionally to be able to view each others’ location.

Rather than focus on the connections between these sellers and the entire population (for an in–depth analysis, see Bakshy, et al., 2010), we examine the interconnections within the Top 100 or Boutique Seller groups. Within each seller group we extract all dyads who bought or sold an item, sent or received a chat message, or bookmarked another user as a friend. We graphically visualize the network, along with attributes such as revenue and chat amount, using GUESS (Adar, 2006).

Figure 4 shows the interconnections between the Top 100 Sellers. The size of the nodes represent the total revenue generated during the month (thicker transaction ties represent more transactions; thicker chat ties edges represent more messages). Our graph also includes multiple edges, allowing us to see when two actors engage in both a transaction and chat, or a reciprocated chat or transaction.

 

Figure 4: Transaction, chat and friend networks of Top 100 Sellers. Green links are transaction networks; blue links are chat networks; red links are friend networks. Node size represents total revenue made in April 2009. Edge weight is number of transactions and chat frequency. When nodes show two lines, it represents different types of relations, or reciprocity within one type of relation
Figure 4: Transaction, chat and friend networks of Top 100 Sellers. Green links are transaction networks; blue links are chat networks; red links are friend networks. Node size represents total revenue made in April 2009. Edge weight is number of transactions and chat frequency. When nodes show two lines, it represents different types of relations, or reciprocity within one type of relation.

 

It is interesting to note the density of the network between the Top 100 Sellers. This “rich club” phenomenon has been observed across a variety of networks, including biological, transportation, and scientific collaboration networks (Colizza, et al., 2006). In fact, Bakshy, et al. (2010) found that among a number of variables relating to sellers’ social interactions, the top predictor of revenue was the revenue of one’s friends. Chat ties are the most prevalent and show that there is a lot of communication between many of these sellers. There are also a substantial number of transaction ties, which shows that Top 100 Sellers frequently sell to each other. Friendship ties are present, albeit less pronounced than the other types of relations. By contrast, the graph for Boutique Sellers is quite sparse. Still, as Figure 5 shows, there are a few cases where Boutique Sellers chat or share a friend list. There is one example of a transaction between two Boutique Sellers.

 

Figure 5: Transaction, chat and social networks of Boutique Sellers
Figure 5: Transaction, chat and social networks of Boutique Sellers.

 

Overall, these networks suggest that Top 100 may be working together in order to increase sales. This might happen by selling items that are reused or modified by other (i.e., wholesale parts), in which frequent communication and friendship can facilitate this exchange. Top 100 Sellers might also be more likely to communicate or maintain friendships because they sell items in the same geographic spaces, acting as a cartel. Finally, the multiple edges in each graph show that some of these relations are reciprocated; two users might sell items to each others, while one user sends a chat messages message and the other engages in a transaction.

Group membership. In addition to the basic communication and friendship networks, we expected to see a connection between economic activity and the groups that sellers belong to. Groups play an important part in Second Life because they help socialize users to the world, bring together users with similar interests, or help promote products or events. In fact, we find that a large portion of the transactions occurs between a seller and a customer who shares at least one group with them: 50.7 percent of transactions for the Top 100 and 49.9 percent of transactions for Boutique Sellers. This suggests that groups perform a vital function for economic activity: sellers can cater to other users who share common interests, or even strategically target certain groups that match their product descriptions.

In order to develop better understanding of group membership in our seller groups, we perform a content analysis of these group genres. After reviewing the various group names and charters of the overall population of users, we devise a coding scheme to represent the basic genres. As depicted in Table 2, groups are focused on lifestyle interests and hobbies (including adult–oriented groups, which make up only a small portion of top seller membership), groups that help sellers produce goods and services such as scripting–oriented groups, or groups focused on music, clubs and DJs and similar entertainment.

 

Table 2: Summary of group categories based on group names and charters.
Group nameDescription
Retail & ScriptingThis category represents groups focused on the design and scripting of retail objects including avatars, clothes and other fashion, furniture, houses, etc. It includes both individual designers (e.g., Mary’s House of Fashion) and groups that share tips on how to create retail objects through Second Life’s scripting language.
Music/Clubs/DJThis category represents groups focused on music groups (e.g., rock, punk, goth, emo or metal bands), DJs and club events, or a single club itself (e.g., Neptune’s Jazzy Dance Club). They might include a fan club for a particular musician, or promotion for a venue.
LifestyleLifestyle groups typically represent social groups (e.g., Friends of Dean), hobby groups (e.g., Aviator Enthusiasts) or adult–oriented groups (e.g., Fuzzies).
Land & RentalsThese groups typically represent a land or community (e.g., Residents of Ibiza Isle), or a group that sells land or rentals (e.g., Antoinette’s Cottage Rentals).
GamesThese groups represent games of skill or chance such as Zyngo, or the variety of venues that house such games (e.g., House of Zyngo; VIP Players Club).
FinancialThese groups typically involve currency exchange (e.g., SLExchange), which allows users to convert the in–game Linden Dollars to real U.S. dollars and vice versa.
Education & ConciergeThese groups represent concierge services for newcomers to Second Life or a particular area, or more academic educational groups (e.g., Scientists Studying Second Life).
AdvertisingThese groups are dedicated purely to advertising products or services and these charters offer a channel for sellers to spam and for buyers to get updated on the latest offerings or events.
CampingThese groups advertise places for Second Life users to ‘camp’ on a specific territory or spot to receive money for populating the area. For example, some clubs pay for campers to dance in order to give a sense that an event is popular and attract other users.
Employment agenciesThese groups represent companies that will staff a building, store or event with other users or bots. For example, agencies can staff a restaurant with employees. Employment agency groups advertise to both owners needing staff, as well as to ‘unemployed’ users looking for work.

 

Figure 6 highlights the distribution of group membership for both Top 100 and Boutique Sellers. Top 100 Sellers belong to a considerable amount of game groups (42.9 percent), which include groups dedicated to Zyngo or other games, VIP and hi–roller clubs that house Zyngo machines, and update and support groups for owners of Zyngo machines or clubs. Many Top 100 Sellers even belong to the same product update group for the makers of Zyngo. In addition to games, popular groups include Retail and Scripting groups (19.7 percent), Land and Rental groups (11.3 percent), Lifestyle groups (14.9 percent), and Music/Clubs/DJ (6.3 percent). There are a handful of Education & Concierge groups, which help users develop and manage their in–world activities, Financial groups, which help users exchange real money for the in–game currency, and Camping groups, which advertise spots where users can ‘camp’ on a particular spot for an extended period of time in exchange for an hourly wage (e.g., dance at a club to help entice others to join).

 

Figure 6: Ratio of group membership types for Top 100 (top) and Boutique Sellers (bottom).
Figure 6: Ratio of group membership types for Top 100 (top) and Boutique Sellers (bottom).

 

 

++++++++++

Discussion and conclusions

The purpose of this study is to provide a description of the seller activity in virtual markets. Given the increasing focus on virtual goods as a viable business model, understanding the economics of popular virtual worlds such as Second Life provides important insights that inform the design of virtual products and delivery mechanisms in online communities. We focus on the most successful sellers to get a better idea of what activities work best. Our findings suggest that many economic and marketing principles associated with real–world products apply to virtual worlds. We find that business cycles exist, but at an hourly level. We find that the best sellers utilize a different set of transactions than a typical boutique seller. We find that geographic concentration occurs even when localization isn’t necessary. And unsurprisingly, we see that social capital plays an important role in the economic behavior of top sellers.

The law of the virtual few. Previous studies suggest that a subset of users tend to contribute the majority of online content (Ling, et al., 2005), and our findings show that a small portion of sellers make up the bulk of sales. Our very top sellers actually make up roughly 40 percent of the total sales revenue in a single month. This is interesting because selling items has a much stronger incentive than contributing a post or answering a question in a voluntary community — free money to sustain one’s Second Life. And while a substantial portion of users did sell at least one item (around 270,000), it pales in comparison to the total users who added at least one friend to their personal networks (over four million). So it is likely that many users are either not interested in selling items or do not understand how, while a handful of savvy users are breaking the bank. Recent news stories highlighting the potential to make real money in Second Life focus on successful sellers who use Second Life as a second home business to make additional cash or attempt to quit their day job. This raises important questions for designers on whether to create interventions to reach a broader set of potential sellers to increase economic activity, or to focus on building additional features for the slice of users who are already succeeding.

The no weekend effect. Cycles, seasons, plateaus, and recessions are all common in economic systems, and retail sectors strategize around weekends and holidays in order to maximize profit. We were surprised to see this weekend effect did not exist in this virtual market. This is partially explained by the lack of temporal constraints that virtual words offer: the freedom to shop 24 hours a day, any day of the week. This implies that virtual markets a potentially lucrative space when normal retail spaces might be constrained by time (for example, during week days and work hours).

The rise of gaming machines. The very best sellers utilize storefronts and automated machines to maximize sales. While gaming machines appear to be big money makers, they point to the potential success of entertainment and amusement in virtual worlds. For example, recent work argues that adolescents are increasingly looking to virtual worlds because of the entertainment factor (Welch, 2010). And this is not limited to three–dimensional worlds: at the time of this publication, games such as Mafia Wars and Farmville dominate Facebook; fifteen of the top 25 highest grossing iPhone apps are games; and, massive multiplayer online games such as World of Warcraft boast over 10 million subscribed users. Our study shows that the most successful sellers are capitalizing on the popularity of games and entertainment.

But what is more interesting is the use of automated machines to provide gaming experiences or other services. Not unlike stand–up arcade games or vending machines, top sellers are leaving machines behind to pull in the most revenue. For example, many of the regions devoted to cash services or currency exchange that we visited contained terminals to let users handle the money exchange on their own. Similarly, there are terminals devoted to shopping catalogues, in which a user can search and buy products without visiting a store at all. Adding automatic virtual goods provides a 24–hour presence for online seller and convenience for an online buyer.

In addition to machines, storefronts and virtual malls are quite prevalent in virtual markets. The focus of these stores is segmented — from teddy bears to BD/SM gear — but they act as searchable, central locations for buyers who are browsing or looking for something in particular. Rather than engage in direct selling through personal networks or solicitations, having a virtual brick–and–mortar store is another form of advertisement. Potential customers entering a new region, or shopping for a different product, become aware of new options. Storefronts also add the potential for cross–selling when an already interested customer searching for a product enters the store. This is amplified when storefronts are concentrated in a single area, as a virtual strip mall or shopping center, or near another event or activity that is popular with users.

Location, location, location. In the physical world, location is everything. From local markets (i.e., a flea market) to commerce centers (e.g., the mall), and beyond (e.g., Wall Street and Hong Kong), a concentration of business leads to success. While the Internet is free from many of the temporal or spatial constraints, in three–dimensional virtual worlds, sellers still benefit from concentrating on a single area. While we find a large distribution of transactions, the ones with the highest volume occur in central locations. One explanation for this is that sellers are utilizing large commerce areas to set up shop; another is that sellers seek out areas or events that regularly draw in a large population of users.

The mystery of (social) capital. Our findings provide further evidence that social capital has an impact on economic performance (Knack and Keefer, 1997), even in online settings. We find much higher frequencies of friendship ties and chat communication among the top sellers, and the networks of the very best sellers reveals an interconnected network of chatter and transactions. There are a couple explanations why the networks of the echelon sellers are so dense. First, these sellers might be providing items or products that other top sellers use or modify for their own products, or acting as intermediaries of wholesale merchandise, one of the most lucrative sector in the U.S. economy (U.S. Census, 2009). Second, these sellers might be networking with other top sellers in order to share tips and tricks for sales and marketing, much like the business networks and chambers of commerce that prevail in regular economies. Third, these sellers might be in cahoots, working together to form and maintain an oligarchy or cartel.

Another important aspect of social capital are found in the types of groups that top sellers join, which often reflect the types of commerce they engage in: retail groups focused on product updates or support; scripting groups to improve product design; music and clubs pages to be aware of popular events; and, gaming groups that notify users where the action is. Joining groups help sellers reach new customers with particular interests and keep existing customers aware of new products or sales. They also keep sellers updated on the latest product updates or techniques to keep their own product lines up–to–date or to better their skills in building products. Together, we find that social interaction and group membership is a vital part of online commerce and the distribution of virtual goods.

In conclusion, our study provides empirical support that the most successful sellers in Second Life reflect many of the characteristics associated with off–line economic activity — daily economic cycles, geographic concentration, and social capital. We also show that the sheer volume of economic activity and the large revenues reinforce recent calls that virtual goods are sure to make a real impact on real-world economies. It is our hope that this descriptive study will prompt continued research into virtual economies. End of article

 

About the authors

David A. Huffaker is a Research Fellow in the School of Information at the University of Michigan and a Visiting Scholar in the Department of Media, Culture and Communication at New York University. For more information and publications, visit http://www.davehuffaker.com.

Matthew Simmons is a PhD student in the School of Information at the University of Michigan.

Eytan Bakshy is a PhD student in the School of Information at the University of Michigan.

Lada A. Adamic is an Associate Professor in the School of Information and the Center for the Study of Complex Systems at the University of Michigan.

 

Acknowledgments

We thank Linden Lab for sharing Second Life data. This work was supported by the Intelligence Community (IC) Postdoctoral Research Fellowship Program, NSF IIS–0746646, and MURI award FA9550–08–1–0265 from the Air Force Office of Scientific Research.

 

Note

1. Castronova, 2001, p. 21.

 

References

E. Adar, 2006. “GUESS: A language and interface for graph exploration,” Proceedings of the SIGCHI conference on Human Factors in computing systems (CHI ’06; Montréal, Canada), pp. 791–800.

E. Bakshy, M. Simmons, D. Huffaker, C. Teng, and L.A. Adamic, 2010. “The social dynamics of economic activity in a virtual world,” paper presented at the Fourth International AAAI Conference on Weblogs and Social Media (ICWSM–10; Washington D.C.); version at http://www-personal.umich.edu/~ladamic/papers/SecondLife/sl-icwsm10.pdf, accessed 23 June 2010.

S. Balasubramanian and V. Mahajan, 2001. “The economic leverage of the virtual community,” International Journal of Electronic Commerce, volume 5, number 3, pp. 103–138.

T. Boellstorff, 2008. Coming of age in Second Life: An anthropologist explores the virtually human. Princeton, N.J.: Princeton University Press.

P. Bourdieu, 2001. “The forms of capital,” In: M. Granovetter and R. Swedberg (editors). The sociology of economic life. Second edition. Boulder, Colo.: Westview Press, pp. 96–111.

C. Cain Miller and B. Stone, 2009. “Virtual goods start bringing in real paydays,” New York Times (6 November), at http://www.nytimes.com/2009/11/07/technology/internet/07virtual.html, accessed 22 February 2010.

E. Castronova, 2002. “On virtual economies,” CESIFO (Center for Economic Studies and Ifo Institute for Economic Research) Working Paper, number 752, at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=338500, accessed 23 June 2010.

E. Castronova, 2001. “Virtual worlds: A first–hand account of market and society on the cyberian frontier,” Gruter Institute Working Papers on Law, Economics, and Evolutionary Biology, volume 2, article 1, at http://www.bepress.com/giwp/default/vol2/iss1/art1/current_article.html, accessed 23 June 2010.

E. Castronova, D. Williams, C. Shen, R. Ratan, L. Xiong, Y. Huang, and B. Keegan, 2009. “As real as real? Macroeconomic behavior in a large–scale virtual world,” New Media & Society, volume 11, number 5, pp. 685–707.http://dx.doi.org/10.1177/1461444809105346

V. Colizza, A. Flammini, M. Serrano, and A. Vespignani, 2006. “Detecting rich–club ordering in complex networks,” Nature Physics, volume 2, number 2, pp. 110–115.http://dx.doi.org/10.1038/nphys209

S. Knack and P. Keefer, 1997. “Does social capital have an economic payoff? A cross–country investigation,” Quarterly journal of economics, volume 112, number 4, pp. 1,251–1,288.

T. Knoop, 2004. Recessions and depressions: Understanding business cycles. Westport, Conn.: Praeger.

H. Konishi, 2005. “Concentration of competing retail stores,” Journal of Urban Economics, volume 58, number 3, pp. 488–512.http://dx.doi.org/10.1016/j.jue.2005.08.005

P. Kotler and G. Armstrong, 2008. Principles of marketing. Twelfth edition. Upper Saddle River, N.J.: Pearson Prentice Hall.

V. Lehdonvirta, 2009. “Virtual item sales as a revenue model: Identifying attributes that drive purchase decisions,” Electronic Commerce Research, volume 9, numbers 1–2, pp. 97–113.

A. Levy and A. Satariano, 2010. “Zynga’s ‘FarmVille’ Facebook games debuts on MSN site,” Bloomberg.com (4 February), at http://www.bloomberg.com/apps/news?pid=20601109&sid=aUJPJRYwSerE&pos=12, accessed 22 February 2010.

K. Ling, G. Beenen, P. Ludford, X. Wang, K. Chang, X. Li, D. Cosley, D. Frankowski, L. Terveen, A.M. Rashid, P. Resnick, and R. Kraut, 2005. “Using social psychology to motivate contributions to online communities,” Journal of Computer–Mediated Communication, volume 10, number 4, at http://jcmc.indiana.edu/vol10/issue4/ling.html, accessed 23 June 2010.

T. Malaby, 2006. “Parlaying value: Capital in and beyond virtual worlds,” Games and Culture, volume 1, number 2, pp. 141–162.http://dx.doi.org/10.1177/1555412006286688

K. Marx, 2003. “Production, consumption, distribution, exchange,” In: D.B. Clarke, M.A. Doel, and K.M.L. Housiaux (editors). The consumption reader. London: Routledge, pp. 251–254.

C. Ondrejka, 2008. “Education unleashed: Participatory culture, education, and innovation in Second Life,” In: K. Salen (editor). The ecology of games: Connecting youth, games, and learning. Cambridge, Mass.: MIT Press, pp. 229–251.

C. Ondrejka, 2004. “Escaping the gilded cage: User created content and building the metaverse,” New York Law School Review, volume 49, pp. 81–101.

J. Peter and J. Olson, 1999. Consumer behavior and marketing strategy. Fifth edition. Boston: Irwin/McGraw–Hill.

M. Porter, 2000. “Location, competition, and economic development: Local clusters in a global economy,” Economic Development Quarterly, volume 14, number 1, pp. 15–34.http://dx.doi.org/10.1177/089124240001400105

V. Reimers and V. Clulow, 2004. “Retail concentration: A comparison of spatial convenience in shopping strips and shopping centres,” Journal of Retailing and Consumer Services, volume 11, number 4, pp. 207–221.http://dx.doi.org/10.1016/S0969-6989(03)00038-9

F. Rothaermel and S. Sugiyama, 2001. “Virtual Internet communities and commercial success: Individual and community–level theory grounded in the atypical case of TimeZone.com,” Journal of Management, volume 27, number 3, pp. 297–312.http://dx.doi.org/10.1016/S0149-2063(01)00093-9

P. Scholten, J. Livingston, and J. Chen, 2009. “Do countercyclical–weekend effects persist in online retail markets?” Electronic Commerce Research and Applications, volume 8, number 4, pp. 174–181.http://dx.doi.org/10.1016/j.elerap.2009.03.003

A. Sen, 1976. “Poverty: An ordinal approach to measurement,” Econometrica, volume 44, number 2, pp. 219–231.http://dx.doi.org/10.2307/1912718

H. Shalit, 1985. “Calculating the Gini Index of inequality for individual data,” Oxford Bulletin of Economics and Statistics, volume 47, number 2, pp. 185–189.http://dx.doi.org/10.1111/j.1468-0084.1985.mp47002006.x

A. Strauss and J. Corbin, 1990. Basics of qualitative research: Grounded theory procedures and techniques. London: Sage.

U.S. Census, 2009. “All sectors: Core Business Statistics Series: Advance Summary Statistics for the United States,” at http://factfinder.census.gov/servlet/IBQTable?_bm=y&-geo_id=&-ds_name=EC0700CADV1&-_lang=en, accessed 10 March 2010.

E. Warner and R. Barsky, 1995. “The timing and magnitude of retail store markdowns: Evidence from weekends and holidays,” Quarterly Journal of Economics, volume 110, number 2, pp. 321–352.http://dx.doi.org/10.2307/2118442

M. Wedel and W. Kamakura, 2000. Market segmentation: Conceptual and methodological foundations. Second edition. Boston: Kluwer Academic.

M. Welch, 2010. “Teens and virtual goods: The fun, useful and affordable luxuries that are driving the virtual economy,” Journal of Virtual Worlds Research, volume 2, number 4, at https://journals.tdl.org/jvwr/article/view/872, accessed 23 June 2010.

L. White, 2002. “Trends in aggregate concentration in the United States,” Economic Perspectives, volume 16, number 4, pp. 137–160.http://dx.doi.org/10.1257/089533002320951019

 


Editorial history

Paper received 5 May 2010; accepted 20 May 2010.


Creative Commons License
This work is licensed under a Creative Commons Attribution–NonCommercial–NoDerivs 3.0 Unported License.

Seller activity in a virtual marketplace
by David A. Huffaker, Matthew Simmons, Eytan Bakshy, and Lada A. Adamic.
First Monday, Volume 15, Number 7 - 5 July 2010
http://firstmonday.org/ojs/index.php/fm/article/view/2977/2569





A Great Cities Initiative of the University of Illinois at Chicago University Library.

© First Monday, 1995-2017. ISSN 1396-0466.